The APL Orchard

« first day (1423 days earlier) ← previous day next day → last day (1528 days later) »

Wezl

23:03

@Marshall here's a full tutorial. maybe I should make a new room for it.

user

23:21

Is there some kind of "scan until" operator in APL? Or a version of the power operator, but where each result is kept in a list and there's a separate state passed around each time?

dzaima

@user not in regular APL

user

But there are other variants with it?

dzaima

@user dzaima/APL has ⍡ that's like ⍣ but gives a list of the results

user

Cool, I'm going to try that out!

dzaima

example (but keep in mind that, just like ⍣, it'll always execute the function at least once, so ⍡ will give a list of at least 2 items)

Dyalog APL

23:23

<moon-child> @user I believe one of the k convergence operators does that

user

I don't know k very well, but I'm going to check that out too, thanks

@dzaima That works for me

dzaima

fwiw this works in regular Dyalog

user

Ah, but that's 6 bytes longer

ngn

@DyalogAPL @ moon-child converge (monad/) and converge-iterate (monad\ ) are analogous to over (aka reduce, dyad/) and scan (dyad\ )

\ preserves intermediate results, / doesn't

Martin Janiczek

"Mastering Dyalog APL" has probably not been updated yet wrt. ⌸ key operator, right? Can't find it in the table of contents

I _think_ I need it:

datapoint_indexes←⍸json_type=4 ∧ p∊⍸'datapoints' in name
5 6 14 20
nearest_node[datapoint_indexes]
2 2 11 17

now ideally I'd like to group the indexes by their nearest node:
2 (5 6)
11 (14)
17 (20)

I was able to do nearest_node {⍺ ⍵}⌸ (json_type=4 ∧ p∊⍸'datapoints' in name) but that destroys the ability to ⍸ later

dzaima

23:36

@MartinJaniczek The actual book hasn't been updated since 2009. here's some effort into updating it, but it doesn't mention ⌸ past its existence

Martin Janiczek

Now that I write that and look at the result, I think I can get it back by adding the nearest_node value and some ⍳ to the right-side array...

dzaima

@MartinJaniczek can't quite infer what "ability to ⍸ later" you want there

Martin Janiczek

When I had the full bool vector there, I could do ⍸ and get 5 6 14 20

But the result of Key splits the bool vector into chunks

(I don't understand Key enough to know whether I can do things in a saner way)

dzaima

@MartinJaniczek (you sure you don't need parentheses around json_type=4?)

Martin Janiczek

uh oh :D

I think I don't need the json_type=4 at all actually, at least for the data I'm testing the program on. Thanks for the push

dzaima

23:42

@MartinJaniczek my first thought is to first just do {⍺⍵}⌸nearest_node and use that Somehow™ (potentially something like {⍺ ⍵ (something[⍵])}⌸?)

@MartinJaniczek there are 2 ways to use ⌸ - monadic and dyadic (and F⌸ ⍵ is equal to just ⍵ F⌸ ⍳≢⍵)

Martin Janiczek

Looks promising, I'm trying to play with the something to get the right bool vector to show up again next to the right ID numbers. But honestly it might be easier for me to go from the other side with the screenshot I've posted

dzaima

(⌸ docs fwiw)

rak1507

'datapoints' in name what is this, python?!

Martin Janiczek

:D blasphemy I know...
equals_string←{(⊂⍺)∊⍨∘⊂⍵}
in←{(⍺ equals_string⊢)¨ ⍵}

I'm sure this is all horribly inefficient

dzaima

@MartinJaniczek equals_string←≡?

in←{(⊂⍺)≡¨⍵}?

Martin Janiczek

23:48

.......... how can I even reply to this

you're right

dzaima

@dzaima alternatively, in←{⍺∘≡¨⍵} or in←{⍵∊⊂⍺}

rak1507

⊢∊⊂⍤⊣ or something

dzaima

also your equals_string doesn't need the ∘, at which point the ⍨ can be removed for {(⊂⍵)∊⊂⍺} which can be golfed to ∊⍥⊂⍨

Martin Janiczek

Not a golfer yet. Just mad at myself that I've tried most of stuff around = but didn't think of ≡

dzaima

@MartinJaniczek with = you could've done {≠/≢¨⍺⍵:0 ⋄ ∧/⍺=⍵}

(assuming the args are of equal length, there's even ∧.= but i personally prefer (∧/=) over that)

rak1507

23:53

I like inner product

Dyalog APL

<moon-child> to be rank-polymorphic, you would need multiple reductions. Which could be done with ∧/⍣≡, but that's circular

Martin Janiczek

Dyalog APL

<moon-child> I think the only alternative is recursion

Martin Janiczek

bingo

dzaima

moon-child: or just ∧/∘,

(btw instead of images you could post paste (shameless self-advertisement); SE chat has a message length limit, but that can be worked around by using multiline messages, but then SE chat stops allowing markdown sooo)

Martin Janiczek

23:57

@dzaima thanks I was missing that!

@dzaima Might be worth adding into the room description next to aplcart and other links